Accuracy and Performance of Continuous
Glucose Monitors in Athletes
Felicity Thomas,* Christopher G. Pretty,* Matthew
Signal,* J. Geoffrey Chase*
*Department of Mechanical Engineering, University of
Canterbury,
Background
Continuous Glucose Monitors (CGMs)
The CGM consists of a pager-like monitoring device that receives information from a sensor inserted
under the skin that detects glucose in the interstitial fluid.
•
Originally designed to help Type 1 diabetics manage blood glucose levels
•
Recently used in the Intensive Care Unit (ICU) and Neonatal Intensive Care Unit (NICU) to detect
hypoglycaemia in at-risk babies
•
CGM accuracy is dependent on Blood Glucose (BG) calibration measurements entered into the device
every four times a day
•
Much more frequent measure of blood glucose (5 minutely) but performance trade offs
(www.medtronic.com) 0 0.2 0.4 0.6 0.8 1 1.2 0 2 4 6 8 Baby 66 CGM data B G ( m m o l/ L ) time (days) CGM BG ref BG
Why?
Optimisation of an athlete’s BG has the potential to
Increase race performance – knowing when and what to eat during racing
Speed recovery – Optimal replacement of glycogen stores
Aid training - as blood glucose can reflect metabolic and inflammatory conditions
However, before these benefits can be realized the accuracy and
performance of CGM devices in active athletes must be evaluated
0 500 1000 1500 2000 2500 3000 3500 4000 0 1 2 3 4 5 6 Time (mins) B lo o d G lu c o s e ( m m o l/ L )
Linear Interpolation Calibration
A Re-calibration algorithm was used to make better use of
the accurate blood glucose measurements.
Background
Linear Regression Calibration
The CGM uses linear regression techniques combined
with smoothing. This is a typical “built-in” method of
calibration.
Blood Glucose
=
slope
* (
electric current
-
offset
)
0 500 1000 1500 2000 2500 3000 3500 4000 0 1 2 3 4 5 6 Time (mins) B lo o d G lu c o s e ( m m o l/ L )
Re-calibrated CGM Trace
passes through BG
calibration measurements
Original CGM Trace
passes near BG
calibration measurements
Recalibration of CGM data
Procedure
Two fasting exercise tests were carried out 3 days apart:
At a later date, ‘fasting sedentary tests’ were carried out.
Rest
Day
+
Fasted
overnight
0
60
120
150
Submaximal
HR zone
(130-140 bpm)
Increase HR
until exhaustion
(190bpm)
1g/kg
Glucose
BG
every
5mins
BG every 10mins
Analysis
𝑀𝐴𝑅𝐷 = 𝑚𝑒𝑎𝑛(𝑎𝑏𝑠
𝐶𝐺𝑀−𝐵𝐺
𝐵𝐺
) ∗ 100
𝑂𝑓𝑓𝑠𝑒𝑡 = 𝐶𝐺𝑀 − 𝐵𝐺
Mean absolute relative difference (MARD) and Offset was calculated between
reference BG measurements collected during the fasting tests and the CGM trace:
These metrics were assessed during:
the exercise or sedentary phase only (0 – 120mins)
Results
Exercise tests
Exercise tests
Sedentary tests
Sedentary tests
CGM Traces
Results – Exercise Performance
During Exercise MARD are equivalent if not better than the performance reported for CGM in
diabetic subjects –
10.8 [8.7 – 16.7] %
median [IQR] or
7.3 [5.4 – 10.9] %
with recalibration
Sensor 1
Sensor 2
Sensor 1
Sensor 2
MARD (%)
exercise +
glucose bolus
7.07
10.1
7.37
12.9
8.74 [7.15 12.2]
MARD (%)
exercise Only
6.64
8
5.01
11.8
7.32 [5.42 10.9]
MARD (%)
exercise +
glucose bolus
8.55
12.6
10.1
16.9
11.4 [8.9 15.8]
MARD (%)
exercise Only
8.73
12.9
8.67
17.9
10.8 [8.69 16.7]
Sensor 1
Sensor 2
Sensor 1
Sensor 2
MARD (%)
exercise +
glucose bolus
18.6
17.8
35.9
34.9
26.4 [17.6 35.2]
MARD (%)
exercise Only
18.1
16.9
37.3
41.5
25.1 [16.9 35.4]
MARD (%)
exercise +
glucose bolus
22.4
18.8
40.8
37.6
30 [21.5 38.4]
MARD (%)
exercise Only
21.6
18.8
43
44.8
32.3 [20.9 43.5]
R
e
ca
li
b
ra
te
d
A
lg
o
ri
th
m
Sedentary Fasting Test 1
Sedentary Fasting Test 2
MEDIAN [IQR]
R
e
ca
li
b
ra
te
d
A
lg
o
ri
th
m
Results – Sedentary Performance
Sensor 1 Sensor 2 Sensor 1 Sensor 2
MARD (%) exercise + glucose bolus 7.07 10.1 7.37 12.9 8.74 [7.15 12.2] MARD (%) exercise Only 6.64 8 5.01 11.8 7.32 [5.42 10.9] MARD (%) exercise + glucose bolus 8.55 12.6 10.1 16.9 11.4 [8.9 15.8] MARD (%) exercise Only 8.73 12.9 8.67 17.9 10.8 [8.69 16.7]
Sensor 1 Sensor 2 Sensor 1 Sensor 2
MARD (%) exercise + glucose bolus 18.6 17.8 35.9 34.9 26.4 [17.6 35.2] MARD (%) exercise Only 18.1 16.9 37.3 41.5 25.1 [16.9 35.4] MARD (%) exercise + glucose bolus 22.4 18.8 40.8 37.6 30 [21.5 38.4] MARD (%) exercise Only 21.6 18.8 43 44.8 32.3 [20.9 43.5] R e ca li b ra te d A lg o ri th m
Sedentary Fasting Test 1 Sedentary Fasting Test 2
MEDIAN [IQR] R e ca li b ra te d A lg o ri th m
Exercise Fasting Test 1 Exercise Fasting Test 2
Sedentary tests obtained worse performance attributed to two main
factors:
•
The reference measurements most likely biased during the sedentary
test due to low apparent skin and leading to BG meters reading lower
than expected values
• Interstitial fluid is not actively pumped like blood. It relies on muscle
movement to circulate and mix.
Result - Bias
There is a consistent positive bias evident, whether it
be exercising or sedentary, or when applying the
recalibration algorithm or the factory algorithm.
Sensor 1 Sensor 2 Sensor 1 Sensor 2
Offset exercise + glucose bolus 0.3 [0.1 0.4] -0.1 [-0.7 0.4] 0.03 [-0.1 0.3] 0.2 [-0.6 0.5] Offset exercise Only 0.3 [-0.002 0.3] 0.3 [-0.1 0.4] 0.2 [-0.05 0.3] 0.5 [0.2 0.7] Offset exercise + glucose bolus 0.4 [0.2 0.5] 0.7 [0.5 0.8] 0.4 [0.1 0.5] 0.7 [-0.3 1.0] Offset exercise Only 0.4 [0.2 0.5] 0.6 [0.5 0.7] 0.4 [0.2 0.5] 0.8 [0.7 1.0]
Sensor 1 Sensor 2 Sensor 1 Sensor 2
Offset exercise + glucose bolus 0.3 [-0.4 1.0] 0.3 [-0.8 0.8] 1.7 [0.9 2.1] 1.6 [0.7 2.1] Offset exercise Only 0.4 [-0.05 1.0] 0.4 [-0.1 0.9] 1.6 [0.6 2.1] 1.8 [0.7 2.3] Offset exercise + glucose bolus 0.7 [-0.03 1.2] 0.4 [-0.3 0.9] 1.9 [1.0 2.5] 1.7 [0.9 2.3] Offset exercise Only 0.7 [-0.4 1.1] 0.5 [-0.3 0.9] 1.9 [0.8 2.5] 1.9 [0.9 2.5] Exercise Fasting Test 2 Exercise Fasting Test 1
Re
ca
libr
at
ed
A
lg
or
it
hm
Sedentary Fasting Test 1 Sedentary Fasting Test 2
Re
ca
libr
at
ed
A
lg
or
it
hm
Limitations
The small data set is a major limitation
of this study, however:
Based on the results of this study an Athlete trial plan
was formed to further test the performance of CGM
devices:
10 fit, healthy adults with a resting heart rate of 60
beats per minute (bpm) or lower
Participants will have 2 Ipro2 CGM devices
(Medtronic Minimed, Northridge, CA, USA) inserted
in to their abdomen at least 24 hours before
Protocol
Very Similar protocol to the first exercise test
Rest
Day
+
Fasted
overnight
0
60
90
150
Submaximal
zone
(2.5W/kg)
Ramp test
until
exhaustion
(20W/5mins)
1g/kg Glucose
BG
every
5mins
BG every 10mins
0.5g/kg Glucose
BG
every
10mins
120
30
Interim Results
Exercise test
Exercise tests
Exercise tests
Exercise tests
Interim Results
8/10 Subjects enrolled so far
Very similar performance between recalibration algorithm and factory algorithm
Very good performance across the board.
Offset no longer evident
MARD
SG1
SG2
SG1 recal
SG 2 recal Median[IQR]
ATH01
11.2
13.4
32.5
12.4
12.9 [11-28]
ATH02
15.2
14.9
9.3
11.5
13.2 [9.8-15]
ATH03
9.0
8.9
6.8
7.7
8.3 [7.0-9.0]
ATH04
12.3
13.8
13.0
ATH05
13.8
11.9
10.8
11.3
11.6 [11-13]
ATH06
12.7
13.8
15.7
17.3
14.8 [13-17]
ATH07
11.1
32.3
22.4
17.9
20.2 [13-30]
ATH08
10.6
14.2
7.5
16.0
12.4 [8.3-16]
Median [IQR]
11.8 [11-14] 13.8 [12-15] 12.3 [7.9-21] 12.4 [11-17]
Conclusions
Good Performance seen with CGM during exercise - Sensors agree well with each other and
reference measurements
During sedentary periods the accuracy of the monitors was reduced - This decrease in
accuracy is likely related to the fact interstitial fluid is not actively pumped like blood. It relies
on muscle movement to circulate and mix.
These result show real promise for using CGM to help optimize BG levels in an athletic
active cohort
These differences in performance also provide insight into how these devices might be more
Future work
Develop Athlete Specific Metabolic Model:
Create Endogenous insulin secretion Model
Create Endogenous glucose production Model
